Fiverr commission for @DM_is_best.
Want a LoRA like this one for your favourite character or original design? Check out my Fiverr gig.
Trained on PonyXL with 37 images, mix of SFW & NSFW so this should generate both.
Trained over 354 epochs with adafactor optimizer & cosine scheduler on 4x NVIDIA Tesla P40 GPUs.
This LORA will not work with SD 1.5 or SDXL. You need to use it with PonyXL or a PonyXL-based checkpoint.
Suggested strength: 0.9
Basic prompt: pichu
Other useful prompts: pokemon
For original model tag feral, for anthro/monster girl tag anthro and source_furry
Reviews are appreciated, please let me know if there's something I could improve.
Pichu is owned by The Pokémon Company, this LoRA is not suitable for commercial uses. I am not responsible for what you do with this LoRA.
Description
FAQ
Comments (4)
Too big for a character LoRA
Depends on your VRAM I guess, I don't like how rank 32 turns out so I set it to 64.
You can have a larger LoRA that accurately reproduces the character or a smaller one that doesn't. I prefer the former.
I have several 100mb character loras that can perfectly reproduce a character. Quit talking gibberish, these aren't checkpoints.
@Drakeni If setting LoRA R to 16 gets the results you want, do that. I don't care about 300MB of extra VRAM, if you do then I'm sorry, my LoRAs aren't for you.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.